Export portable Live evidence bundles

AI-assisted implementation under maintainer direction.

tests/framework-fixtures/README.md

@@ -120,3 +120,38 @@ When `preActions` is omitted, steer smoke inherits `runtime.preActions` to revea
 | `nuxt-csp/` | Nuxt `routeRules` with literal CSP header in `nuxt.config.ts`. CSP shape `append-string`. |
 
 Add new fixtures by cloning a directory, swapping files, and updating `fixture.json`.
+
+## External quality-eval fixtures
+
+The public Live benchmark can execute a fixture owned by another repository
+without copying its task corpus or rubric into Impeccable:
+
+```sh
+bun run bench:live -- \
+  --fixture-dir=/absolute/path/to/private-fixture \
+  --agent=codex \
+  --action=bolder \
+  --iterations=1 \
+  --evidence-bundle=/absolute/path/to/output-bundle
+```
+
+An external fixture has the same shape as a directory in this folder:
+`fixture.json`, `gitignore.txt`, and `files/`. Use the optional
+`evidenceCapture` block in `fixture.json` for rubric-free capture metadata:
+
+```json
+{
+  "evidenceCapture": {
+    "captureSelector": "section.case-study",
+    "viewport": { "width": 1440, "height": 1080 },
+    "action": "bolder"
+  }
+}
+```
+
+The bundle contains `report.json`, the original capture, each progressively
+delivered variant capture, geometry/overflow facts, hashes, and timing data.
+It deliberately cannot run `--judge-rendered`; comparative rubrics, private
+fixtures, human calibration, and quality decisions belong in the consuming
+evaluation harness. The normal public E2E suite remains responsible for Live
+protocol, framework, source-commit, cleanup, and recovery correctness.

tests/live-benchmark.test.mjs

@@ -9,6 +9,7 @@ import {
   deriveJournalGenerationMetrics,
   durationBetween,
   parseLiveBenchmarkArgs,
+  resolveLiveBenchmarkPaths,
   summarizeRuns,
 } from '../scripts/lib/live-benchmark.mjs';
 
@@ -27,6 +28,36 @@ describe('live benchmark metrics', () => {
     });
   });
 
+  it('resolves an external fixture into a portable rubric-free evidence bundle', () => {
+    const paths = resolveLiveBenchmarkPaths({
+      fixtureDir: '../impeccable-evals/fixtures/live/tidewater',
+      evidenceBundle: '/tmp/live-tidewater',
+    }, {
+      root: '/workspace/impeccable',
+      fixturesDir: '/workspace/impeccable/tests/framework-fixtures',
+    });
+    assert.deepEqual(paths, {
+      fixtureName: 'tidewater',
+      fixtureDir: '/workspace/impeccable-evals/fixtures/live/tidewater',
+      fixtureOrigin: 'external',
+      evidenceRoot: '/tmp/live-tidewater',
+      artifactRoot: '/tmp/live-tidewater',
+      outputPath: '/tmp/live-tidewater/report.json',
+    });
+  });
+
+  it('keeps evidence bundles atomic and unambiguous', () => {
+    const options = { root: '/repo', fixturesDir: '/repo/tests/framework-fixtures' };
+    assert.throws(
+      () => resolveLiveBenchmarkPaths({ evidenceBundle: 'bundle', artifacts: 'shots' }, options),
+      /replaces --artifacts/,
+    );
+    assert.throws(
+      () => resolveLiveBenchmarkPaths({ evidenceBundle: 'bundle', output: 'report.json' }, options),
+      /writes report.json itself/,
+    );
+  });
+
   it('derives production worker phases from the durable session journal', () => {
     const metrics = deriveJournalGenerationMetrics({
       generationTimings: {

tests/live-e2e/session.mjs

@@ -32,8 +32,7 @@ export { SCRIPTS_DIR, FIXTURES_DIR, REPO_ROOT };
 // Stage
 // ---------------------------------------------------------------------------
 
-export function stageFixture(name, fixture) {
-  const fixtureRoot = join(FIXTURES_DIR, name);
+export function stageFixture(name, fixture, { fixtureRoot = join(FIXTURES_DIR, name) } = {}) {
   const gitignore = readFileSync(join(fixtureRoot, 'gitignore.txt'), 'utf-8');
 
   const tmp = mkdtempSync(join(tmpdir(), 'impeccable-e2e-'));
@@ -216,6 +215,7 @@ export async function stopDevServer(child) {
  * @param {object} opts
  * @param {string} opts.name              fixture name
  * @param {object} opts.fixture           fixture.json contents
+ * @param {string=} opts.fixtureRoot      fixture directory; defaults to the public framework fixture tree
  * @param {import('playwright').Browser} opts.browser   shared browser instance
  * @param {object} opts.agent             VariantAgent (defaults to fake)
  * @param {object|function=} opts.wrapTarget live-wrap target or event mapper
@@ -228,6 +228,7 @@ export async function stopDevServer(child) {
 export async function bootFixtureSession({
   name,
   fixture,
+  fixtureRoot,
   browser,
   agent,
   wrapTarget,
@@ -244,7 +245,7 @@ export async function bootFixtureSession({
   const runtime = fixture.runtime;
   if (!runtime) throw new Error(`fixture ${name} has no runtime block`);
 
-  const tmp = stageFixture(name, fixture);
+  const tmp = stageFixture(name, fixture, { fixtureRoot });
   let live;
   let dev;
   let agentAbort;

tests/live-rendered-quality.test.mjs

@@ -47,6 +47,26 @@ describe('Live rendered quality judge', () => {
     assert.equal(context.safeContext.componentRoles.ActionLink, 'Quiet outlined control');
   });
 
+  it('prefers rubric-free evidence capture settings for external harnesses', () => {
+    const context = buildRenderedReviewContext({
+      fixture: 'private-fixture',
+      fixtureConfig: {
+        runtime: { pickSelector: '.picked' },
+        evidenceCapture: {
+          captureSelector: '.selected-section',
+          action: 'bolder',
+        },
+        renderedQuality: {
+          captureSelector: '.public-smoke-only',
+          reviewFocus: 'Must not leak into the evidence contract.',
+        },
+      },
+    });
+    assert.equal(context.captureSelector, '.selected-section');
+    assert.equal(context.action, 'bolder');
+    assert.equal(context.safeContext.reviewFocus, '');
+  });
+
   it('requires every expected rendered variant to pass the strict score floor', () => {
     const result = parseRenderedJudgeResult(JSON.stringify({
       variants: [